96 vs 88.2 and down sampling (again?)

[houser]

This must have been covered here in the past, but I must be searching for the wrong terms here on Gearslutz?
very few posts with the numbers in them...
.post a link to subject if it's here please?
Ah well,
I am recording a choir next week in a nice big hall and want to record at the best SR for future use.
First time recording at higher bit debts in a sensitive situation for myself...
PT HD 192 IO
RME Mic Pre RME: Micstasy
going into the 192 va AES/EBU.
Expensive german Mics that shoudl do the job...
Planning to use the Digi sync too,,
So:
Should I be using 96 or 88.2 kHz?
The first use for this recording will be a CD, so I will SR convert the final mixes very soon.
What do the downsample sound like in PT 7 from 96vs 88.2 to 44.1?
I would assume 88.2 is a good idea for the CD but are there obvious ifs and buts?

Any other things I need to consider?
Grateful for any hints...

TIA and regards
Janne A.

[David Rick]

In theory, 88.2k and 96k should re sample equally well to 44.1k. In practice they often don't, because many sample-rate conversion routines are badly written. There are two issues:

1) Does the software do a good sample rate conversion?

2) Is it efficiently written for the non-factor-of-two situation?

I use R8BrainPro. While it does a very good conversion from 96k to 44.1k, it does it very slowly because the author apparently didn't know how to do it using polyphase filtering. Converting from 88.2k to 44.1k goes much faster.

Consequently, I tend to record at 88.2k when the project is destined for CD, and at 96k when it is destined for DVD.

David L. Rick
Seventh String Recording

[boojum]

Originally Posted by David Rick

In theory, 88.2k and 96k should re sample equally well to 44.1k. In practice they often don't, because many sample-rate conversion routines are badly written. There are two issues:

1) Does the software do a good sample rate conversion?

2) Is it efficiently written for the non-factor-of-two situation?

I use R8BrainPro. While it does a very good conversion from 96k to 44.1k, it does it very slowly because the author apparently didn't know how to do it using polyphase filtering. Converting from 88.2k to 44.1k goes much faster.

Consequently, I tend to record at 88.2k when the project is destined for CD, and at 96k when it is destined for DVD.

David L. Rick
Seventh String Recording

OK, I am using SAM8 and have been assuming that since all the math is done in binary it should not make a difference. You are saying that there are many ways to do this and that all of them are not equally good. I have been doing 24/48 and downsampling. I suppose I could go to a higher sampling rate, but wonder in SAM8 if the rate need be a "divisible by two" number?

Can you help on this as it applies to SAM8??

[mds]

By SAM8 are you referring to Samplitude 8, or something else? Samplitude has very good SR conversion...

I've done these types of projects in the past at 88.2k for CD projects and 96k for DVD projects as well. All things being equal it shouldn't matter, but this approach seems like the safest one to me.

[boojum]

Originally Posted by mds

By SAM8 are you referring to Samplitude 8, or something else? Samplitude has very good SR conversion...

I've done these types of projects in the past at 88.2k for CD projects and 96k for DVD projects as well. All things being equal it shouldn't matter, but this approach seems like the safest one to me.

Yes, Samplitude 8 SE. I like safer/safest. I tried the other way and it is not as good.

[David Rick]

Originally Posted by boojum

You are saying that there are many ways to do this and that all of them are not equally good. I have been doing 24/48 and downsampling. I suppose I could go to a higher sampling rate, but wonder in SAM8 if the rate need be a "divisible by two" number?

Can you help on this as it applies to SAM8??

In Samplitude 8, the quality of the SRC depended on what menu you accessed it from. Even though you got an identical dialog to set the SRC quality, the results were not always the same. If I recall, the best way was to do an explicit export with SRC to a new file, or to invoke SRC in destructive editing. Still, the best SRC algorithm wasn't up to what you could do in the free version of R8Brain, so I switched to that and then upgraded to R8BrainPro.

In Samplitude 9, a significantly better SRC algorithm was added. (I don't know if this made it into Sam 9SE or not.) Also, the inconsistent SRC behavior was fixed so it doesn't matter what menu you invoke it from.

I'm sorry to admit that I haven't yet gotten around to doing a direct comparison between the new Samplitude 9 SRC and R8BrainPro. Out of habit, I tend to use the latter for most serious stuff, but now I'm not afraid to use Sam9's built-in SRC if I'm in a time crunch. I haven't really noticed whether it is faster at x2 ratios or not.

David L. Rick
Seventh String Recording

[John Brook]

What would you gain by using 96 vice 88.1? If none of your signal above 22.05khz is going to be used on the CD???

[AudioFields]

88.2 / 2 = 44.1 = CD
96 / 2 = 48 = DVD

this makes the most sense to me without trying to understand polyphasing. Id stick with what makes sense. Unless you're into math audio.

[larsfarm]

Originally Posted by AudioFields

88.2 / 2 = 44.1 = CD
96 / 2 = 48 = DVD

this makes the most sense to me without trying to understand polyphasing. Id stick with what makes sense. Unless you're into math audio.

It seems to make sense, but it doesn't when you scratch the surface. It seems that you could just drop every other sample or perhaps something similar, but you can't because then you get aliasing. You must asume that there is some frequency content up there. That must be removed without leaving traces. Since you need to remove all frequency content above the target you need to do the very same thing in both cases.

L

[houser]

Hohn,
If that question is for me, i wrote there is planned intended future use at higher bitdebths. The CD is just the first media it will be used on.
best
Janne A.

Originally Posted by John Brook

What would you gain by using 96 vice 88.1? If none of your signal above 22.05khz is going to be used on the CD???

[AudioFields]

Originally Posted by larsfarm

It seems to make sense, but it doesn't when you scratch the surface. It seems that you could just drop every other sample or perhaps something similar, but you can't because then you get aliasing. You must asume that there is some frequency content up there. That must be removed without leaving traces. Since you need to remove all frequency content above the target you need to do the very same thing in both cases.

L

well put...and well understood. I guess this will have a greater impact with the coming of uncompressed audio via BluRay and HDDVD.?.?.?.?

[MichaelPatrick]

Originally Posted by AudioFields

88.2 / 2 = 44.1 = CD
96 / 2 = 48 = DVD

I think Audiofields has it right. If there's a good chance the material will end up oo a DVD or visual media production, then a 96 is the best source rate. Just use decent SRC to get it down to a CD.

If, however, your target is CD only, then 88.2 is the way to go. I record most things at 88.2. R8Brain Pro works well for me converting rates when needed.

[big country]

Originally Posted by AudioFields

88.2 / 2 = 44.1 = CD
96 / 2 = 48 = DVD

this makes the most sense to me without trying to understand polyphasing. Id stick with what makes sense. Unless you're into math audio.

bingo

[Remoteness]

Originally Posted by MichaelPatrick

I think Audiofields has it right. If there's a good chance the material will end up oo a DVD or visual media production, then a 96 is the best source rate. Just use decent SRC to get it down to a CD.

If, however, your target is CD only, then 88.2 is the way to go. I record most things at 88.2. R8Brain Pro works well for me converting rates when needed.

Yes, David from Seventh String Recording addressed this in his earlier post above...

He stated, "Consequently, I tend to record at 88.2k when the project is destined for CD, and at 96k when it is destined for DVD."


This makes a lot of sense.

[John Brook]

John,
If that question is for me, i wrote there is planned intended future use at higher bitdebths. The CD is just the first media it will be used on.
best
Janne A.

The bit rate is 88.2 or 96 khz; the bit depth is 24 bits per word, in both cases. As far as I understand it, higher bit rates cannot be translated to greater bit depth. I'm open to instruction on this matter, as always!

[big country]

so in this same theory (1/2 1/2)
shouldn't you record at 32 BIT?
for CD audio
44.1 16bit
88.2 32 bit


it makes sense
just for the simplicity of math

or would you be better of recording at 8 bit
rather than 24 bit

I wonder if you could run 8 bit 88.2
on a CD

and would it sound better?
my simple theory tells me yes

wouldn't the binary code be close to equal in length
16 /44.1
8 /88.2

[AudioFields]

Originally Posted by big country

so in this same theory (1/2 1/2)
shouldn't you record at 32 BIT?
for CD audio
44.1 16bit
88.2 32 bit


it makes sense
just for the simplicity of math

or would you be better of recording at 8 bit
rather than 24 bit

I wonder if you could run 8 bit 88.2
on a CD

and would it sound better?
my simple theory tells me yes

wouldn't the binary code be close to equal in length
16 /44.1
8 /88.2

i think this "theory/rule" only applies when down converting the sample rate. I "think" the rule for bit would be....the higher the better, no matter what. Does this sound right to anyone else?

[big country]

I would think that sample rate is more important than bit depth
in sound quality

[AudioFields]

Originally Posted by big country

I would think that sample rate is more important than bit depth
in sound quality

you may have to excuse my ignorance on this one but here it goes....Does bit directly control what frequencies are in the mix. example...16 bit will only have frequencies up to 16khz...and 24 bit will only have up to 24khz.

I know there are some dudes doing some research on what effect the higher frequencies (that we cant hear with the human ear) do to the mix via psychoaccoustics.

[John Brook]

I believe bit depth, ie 12, 16, 20, or 24 bits control the dynamic range of your recording. 12 bits gives you about as much dynamic range as old LPs and FM radio; 16 bits gives you roughly 100 db of dynamic range, and 24 bits bumps it up to somewhere are 124 db (theorectically). I think this issue has been hashed on the high end forum recently.

Where as the recording frequency controls the highest possible audio signal which can be reproduced. 44.1khz was selected because it can produce a pure sine wave at 22.0hkhz, which is above the range of human hearing. Dan Lavry has an excellent technical paper which most of us can follow, at his Lavry website.

[big country]

Originally Posted by AudioFields

you may have to excuse my ignorance on this one but here it goes....Does bit directly control what frequencies are in the mix. example...16 bit will only have frequencies up to 16khz...and 24 bit will only have up to 24khz.

I know there are some dudes doing some research on what effect the higher frequencies (that we cant hear with the human ear) do to the mix via psychoaccoustics.

Anti-Alias Filter: Slide 2 of 6

would be nice to ditch the anti alias filter



think of sample rate as frames pr.second on film
then think of bit rate as the # of pixels or definition

[ghellquist]

Very short and simplified theory, trying to set some misconceptions earlier right.

Sample rate : how often samples are taken. One typical example is 44100 times a second on a CD, also written as 44.1kHz. Half the sample rate sets the theoretical upper limit on frequencys that can be reproduced. For a 44.1k CD that would be 22050Hz. Actual limit is always lower.

Bits / sample : the number of bits used for each sample. Typically 16 or 24, CD-s use 16. The number of bits sets the theoretical maximum distance between the signal and the noise floor. This measure known as SN ratio is proportional to about 6dB per bit *). 16 bit will give a maximum of 96dB, 24 bits a maximum of 144dB. Best current AD converters used for audio comes in at around 20 bits, maybe one or two more if you leave mass produced and go for silly money. There can be other reasons to work with floating numbers at 32 bits or more, one is that subsequent processing is done more easily.

Bit rate : the number of bits per second. For a CD that is 16 x 44.1 x 2 or around 1411 k bits per second. Mp3 coding can often sound good at 128k bits per second.

Gunnar

*) This is a simplification, but for the discussion it is accurate enough.

[big country]

are the ears smarter than the eyes

I believe so

[big country]

SACD uses a very different technology from CD and DVD-Audio to encode its audio data, a 1-bit delta-sigma modulation process known as Direct Stream Digital at the very high sampling rate of 2.8224 MHz. This is 64 times the sampling rate used in Compact Disc Digital Audio, which specifies 44.1 kHz at a resolution of 16-bit.

so by the theory above their is virtually no dynamic range?

[Oroz]

Originally Posted by John Brook

I believe bit depth, ie 12, 16, 20, or 24 bits control the dynamic range of your recording. 12 bits gives you about as much dynamic range as old LPs and FM radio; 16 bits gives you roughly 100 db of dynamic range, and 24 bits bumps it up to somewhere are 124 db (theorectically). I think this issue has been hashed on the high end forum recently.

Just to get a little more precise on the topic you get 6dB of dynamic range for every bit:

1 bit = 6 dB of dynamic range
16 bits = 96 dB of dynamic range
24 bits = 144 dB of dynamic range

Where as the recording frequency controls the highest possible audio signal which can be reproduced. 44.1khz was selected because it can produce a pure sine wave at 22.0hkhz, which is above the range of human hearing. Dan Lavry has an excellent technical paper which most of us can follow, at his Lavry website.

Don't forget the Nyquist Theorem which states that "in order to digitally encode the desired frequency bandwidth, the selected smaple rate must be at least twice as high as the highest frequency to be recorded"

[big country]

SACD does not have a head room of 6db

does any body know the Watts scale

lieqwest smiqest

some high end enternianment
YouTube - the touch

[ghellquist]

Well, I told you 6dB per bit is a simplification.

SACD though is something different, it does not do traditional one bit sampling and the old 6dB per bit simply does not apply *). A SACD disc has a bit rate of about 5600 kBits per second (also written as 5.6Mbits per second), about four times a normal CD.

As always there are tradeoffs to be made -- moving the maximum frequency upwards increases noise. SACD has been designed by the engineers to give about 120dB SN ratio (equivalent to about 20 bits in a normal system, which is about what the best AD converters give you) and often bandwidht limited to about 50kHz. This is roughly equivalent to what you would get from a 120kHz 20bit normally sampled system. This might make you believe that a 192kHz 24 bit system is better than SACD, and well, the math says it is. The reality, as always, takes its own twist and turns.

Gunnar

*) It takes some rather sophisticated mathematical tools to prove the 6dB per bit figure of normal sampling. I absolutely cannot do the math myself.